Strath Tutor © the development and evaluation of a learning-by-browsing system on the Macintosh

Abstract An account is given of the major influences on the design of StrathTutor, a leaming-by-browsing tutorial system for the Apple Macintosh computer. In CAL programming, explicit links between frames in a tutorial represent a limiting and expensive approach to authoring and Al techniques are widely expected to point the way to future progress. Intelligent CAL appears to be concentrating on problems thought to be amenable to the expert system approach, i.e. highly procedural and well analysed tasks. Developments in technology are tending to force the pace. Interactive laser-discs and CD-ROM's allow fast random access to extensive graphic and textual data. Learning will come to include interaction with complex data structures by means of browsing, scanning, wandering and exploring. Navigation through the data, and ease of authoring, will require more sophisticated forms of knowledge representation than indexation, including semantic networks, production rules and error libraries. Ultimately, the issue of learner control will require to be confronted, basically because the best possible model of the learner's state of knowledge is that residing in the learner. In StrathTutor, knowledge is represented by attribute coding of each frame, influenced by theoretical work from two apparently unrelated areas: the computational model of human memory MINERVA 2 and expert systems based on pattern recognition. Information in the form of frame contents and as attributes is separate from the tutorial shell which conducts the interaction with the learner, with the aid of the Macintosh user interface. STmaker operates in “authoring” mode allowing the composing of new tutorials, and STviewer replays a trace of a student's path through a tutorial. The menu choices in StrathTutor are described. Evaluation of a system such as Strath Tutor is being attempted in three ways: comparison of effectiveness of learning with traditional forms of instruction, cognitive analysis while the student learns, and formal and informal assessments in the hands of both teachers and students. Groups using Strath Tutor have shown significant learning. The inherent difficulties encountered have led to better modes of experimentation being devised, and to suggested modification of Strath Tutor itself.